Humanoid robot stocks and ETFs: who's actually public. Offbrand Human

This is not financial advice. I track robot prices, not stock prices. Nothing here is a recommendation to buy, sell, or hold any security. Do your own research and talk to a licensed advisor before investing.

You cannot buy stock in most of the humanoid robot companies you have heard of, because they are private. That is the short answer to "humanoid robot stocks," and I know it is not the answer anyone typing that phrase wants. 1X, Figure, Apptronik, and Booster are not on any exchange. Unitree is the exception in motion: it launched a Shanghai STAR Market IPO in late July 2026, a raise of about $620 million per Caixin's coverage, with subscriptions opening August 10, 2026. Even then, STAR Market shares are hard for ordinary US retail investors to buy directly. The rest of the public exposure that exists is either a giant where humanoids are a rounding error, or a supplier selling picks and shovels to everyone.

Unitree's white and black G1 humanoid walking across a paved shopping plaza in daylight, past a delivery rider in a yellow jacket and an older man in a checked coat who are both turned toward it; Unitree is the only company on this page with a listing actually in progress. — Not a studio sweep: the G1 out on a plaza in Unitree's own marketing still, two people stopped mid-errand to watch it walk. Photo: Unitree

Public vs private: the map

Company	Flagship	Public or private?	How you'd get exposure
Tesla	Optimus	Public (Nasdaq: TSLA)	Optimus is an in-house Tesla program, not a separate company
XPeng	Iron	Public (NYSE: XPEV)	Iron is an in-house division of the carmaker
NVIDIA	(compute supplier)	Public (Nasdaq: NVDA)	Sells the AI compute inside these robots (NEO runs Nvidia Jetson Thor); an Apptronik partner
1X Technologies	NEO	Private	No public shares
Figure AI	Figure 03	Private	No public shares
Apptronik	Apollo	Private	No public shares
Unitree	G1 / R1 / H2	Private, IPO in progress	Shanghai STAR Market IPO underway: subscriptions open August 10, 2026. STAR shares are hard for ordinary US retail investors to buy directly
Booster Robotics	T1 / K1	Private	No public shares
UBTech	U1	Public (HKEX: 9880)	Named by TrendForce as a companion-humanoid maker

Eight stacked rungs headed 'Humanoid robot stocks: the map', seven of them carrying a product photo. Figure AI, no shares, flagship Figure 03, not listed on any exchange, shown as a grey knit-suited torso with both arms. Apptronik, no shares, flagship Apollo, NVIDIA is a listed partner so NVDA trades while Apptronik does not, shown as a white Apollo head with an A1 chest display. 1X and Booster, no shares, the 1X NEO plus the Booster T1 and K1, with the T1 pictured mid-stride over a soccer ball on a grass pitch: you can put a deposit on a NEO but cannot buy a share of 1X. Unitree, $620M, Shanghai STAR Market IPO in progress with subscriptions opening Aug 10, 2026 and STAR shares hard for US retail to buy directly, shown as a black Unitree H1 humanoid holding its stance on a concrete plaza while a person's foot pushes at its hip in a balance test. The public names, 4 tickers: Tesla TSLA and XPeng XPEV are carmakers, NVIDIA NVDA sells the compute, UBTech is HKEX 9880, shown as a line of UBTech U1 units posed on podiums at the maker's launch event. KOID on Nasdaq, Tesla 1.72% of the humanoid-branded ETF with NVIDIA 2.13% and UBTech 2.04%, $296.9M in assets on Aug 3, 2026, and the graphic's one text tile reading 0.69% net fee. Jibo, $73M raised and assets sold off by 2020, shown as the white tabletop robot on black. Moxie, late 2024, maker Embodied shut down when a funding round fell through, shown standing on a striped rug.

"Is Optimus part of Tesla?" Yes, and that is the catch

There is no Optimus stock. Optimus is a program inside Tesla, so buying "Optimus" means buying Tesla, a company whose value is overwhelmingly cars and energy, with the robot as a bet on top. And it is an unproven bet: as of January 2026, Elon Musk said no Optimus robots are doing useful work at Tesla, and the V3 reveal has slipped repeatedly through 2026. There is still no consumer price or checkout. So "tesla optimus stock" is really a question about how much of Tesla's price you think its robot ambition deserves, which is a Tesla question, not a robot question. I track the robot side of it on the Optimus tracker.

A Tesla Optimus standing full length against a black studio background in Tesla's own photo, with a brushed silver chest plate, bare silver actuator arms and legs, and a blank glossy black faceplate; this is the robot people mean when they search for a stock, and the only ticker it lives under is TSLA, the whole car company. — An earlier Optimus build in Tesla's own studio shot: no bodywork over the limbs, a round star medallion at the hip, and nothing on it that trades. Photo: Tesla

It helps to see what that bet looked like when it was being sold. In September 2024, a clip posted by @tsarnick summarized the pitch as Optimus "produced at the scale of a million units a year and at the cost of $20K each" by 2030. Hold that next to the January 2026 admission above, and the gap between the two is the part of Tesla's price that is a robot story rather than a robot business.

Clip of Elon Musk projecting Optimus at a million units a year and $20K each by 2030 (@tsarnick, 2024-09-10)

A timestamped version of the Optimus projection that a TSLA position is implicitly paying for, so the January 2026 status has a dated number to be measured against rather than a vibe.

The same logic applies to XPeng: Iron is a division of an EV company, currently used on XPeng's own production lines, with mass production targeted for the end of 2026. You would be buying a carmaker with a robotics side project, not a robot company.

The "arms dealer" angle

The one place a humanoid boom shows up in a public company without you betting on any single robot surviving is the supply chain. Every credible forecast (Goldman, Morgan Stanley, Bank of America) agrees the near-term humanoid volume is industrial, and all of it needs compute, actuators, and sensors. NVIDIA (Nasdaq: NVDA) is the clearest example: its Jetson Thor is the brain inside the 1X NEO, a pairing 1X showed off around NVIDIA's GTC 2026 event, and NVIDIA appears as a partner on Apptronik's site. That is the "sell shovels in a gold rush" thesis. It is also not specific to humanoids, which is exactly why it is safer and less exciting.

The keynote where NVIDIA made the shovels-seller case in its own words, robotics partners included; useful for judging how much of the pitch is product and how much is stagecraft.

The "humanoid robot ETF" problem

People search "humanoid robot etf" expecting a fund of pure-play humanoid makers. Two humanoid-branded funds now exist, and neither is that. The KraneShares Global Humanoid Robotics and Physical AI Index ETF (Nasdaq: KOID) is passive, tracks an index, launched June 4, 2025, and held about $296.9 million in assets as of August 3, 2026 per the fund page, with a 0.69% net expense ratio (0.79% gross). The Roundhill Humanoid Robotics ETF (Cboe: HUMN) is actively managed, launched June 26, 2025, with a 0.75% expense ratio.

The KraneShares KOID fund page, headed 'KraneShares Global Humanoid Robotics and Physical AI Index ETF', with copy saying KOID is the first U.S. listed humanoid robotics ETF and seeks to track the MerQube Global Humanoid Robotics and Physical AI Index, an Expense Ratio block reading 0.79% (Gross) and 0.69% (Net), a NAV of $38.65 as of 08/05/2026, a NAV total return of YTD 15.61% as of 07/31/2026, and a Fund Details panel listing Ticker KOID and Primary Exchange NASDAQ. — The humanoid-branded fund describes itself in its own first paragraph as an index tracker, and prices itself at 0.69% net, 0.79% gross. The holdings and the asset total sit further down that page, below this frame. Source: kraneshares.com/etf/koid/.

Neither fund can fix the structural problem, because the pure plays (1X, Figure, Apptronik, Booster) are private, and a fund cannot hold what does not trade. What they hold instead are public proxies and suppliers, at small weights. In KOID, per the fund's own holdings table as of August 3, 2026, NVIDIA is 2.13% of the fund, UBTech is 2.04%, and Tesla is 1.72%. That is the humanoid-branded fund. HUMN's top holdings also lean on UBTech and Tesla, per the fund page's June 30, 2026 display.

Beyond those two, what fills the search results are broad robotics and automation ETFs, which hold industrial-automation, semiconductor, and machine-vision companies: the biggest, Global X's BOTZ (Nasdaq), held $3.40 billion in net assets as of August 3, 2026, with Keyence, ABB, NVIDIA, and Fanuc up top and no humanoid pure-play anywhere in its top ten. If you buy any of these expecting concentrated 1X-and-Figure exposure, you will not get it.

The reason the pure plays stay private (and the graveyard footnote)

Private companies raise from venture investors precisely so they can burn money for years before a product proves out, which is the whole humanoid situation. It produces a strange retail moment: you can put down a deposit on a 1X NEO today, but you cannot buy a single share of the company that makes it.

1X's NEO, in its cream knit suit with the lit ring on its face, holding a feather duster up to a shelf of decorative plates in a warm wood-paneled living room, from the maker's press kit; this is the product you can put a deposit on while none of the equity behind it trades anywhere. — 1X's own set dressing: family photos, a matched set of bound volumes, a trailing houseplant, and one robot with a duster. Photo: 1X Technologies

The category's track record should also temper any "get in early" instinct. The consumer-robot graveyard is full of well-funded companies that never made it to an exit. Anki raised about $182.5 million and went bankrupt in April 2019. Jibo raised about $73 million (the commonly cited Crunchbase figure) and had its assets sold off by 2020. Moxie's maker, Embodied, shut down in late 2024 when a funding round fell through. Being early to a company is not the same as being early to a winner.

Jibo, the small white tabletop robot with a round black screen for a face, sitting on a glass coffee table while a smiling man in a green sweater leans toward it from the couch, a lit Christmas tree behind him, in the company's own store photography; this is one of the well-funded companies that never reached an exit. — Jibo Inc.'s own holiday lifestyle shot, from the store it no longer has. Photo: Jibo Inc.

What you would actually own

The humanoid names with real momentum (1X, Figure, Apptronik, Booster) are private. You cannot buy them on an exchange today. Unitree is the one mid-exit: its Shanghai STAR Market IPO is in progress (subscriptions open August 10, 2026), and STAR shares are hard for ordinary US retail investors to access.
The public tickers people reach for (Tesla, XPeng) are diversified companies where the humanoid is a side bet, not the business.
The cleanest public exposure to the trend is suppliers, and it is diffuse by design.
Humanoid-branded ETFs now exist (KOID and HUMN), but the private pure plays are not in them; even in KOID, Tesla is a 1.72% position (August 3, 2026, per the fund page). The rest of the funds under that search are broad robotics baskets.

Whatever you decide to do with that, the thing on the other side of every ticker on this page is a car company, a chip supplier, or an index that cannot hold the pure plays because they do not trade. For the product-side reality that underpins all of it, see how much a humanoid robot really costs and the best humanoids ranked by what you can verify.

Not advice, just the ownership reality.